Summarization on PubMed 205w (test)
45.1ROUGE-1Llama-3.1-8B AR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Llama-3.1-8B ARDecoding Paradigm=Autoregressive, Prompt Template=chat-template2026.05 | 45.1 | 16.5 | 24.5 | |
| Qwen2.5-7B ARDecoding Paradigm=Autoregressive, Prompt Template=chat-template2026.05 | 42.4 | 14.2 | 22.6 | |
| LLaDANFE=64, Decoding Paradigm=Diffusion, Prompt Template=chat-template2026.05 | 42.2 | 15.8 | 23.2 | |
| DSL-LLaDA-SDENFE=64, Decoding Paradigm=Diffusion, Prompt Template=chat-template2026.05 | 39.5 | 14.9 | 22.8 | |
| XDLM Base + xDMNFE=64, Decoding Paradigm=Diffusion, Prompt Template=plain prompt, Sampler Config=k1=0.12026.05 | 18.9 | 4.9 | 12.4 |